st: VS: predictors when using xtmelogit

[Søren Teglgaard Jakobsen - KREVI <STJ@krevi.dk>]

I am working with hierarchical ordered data where the variable of interest is binominal, measuring whether or not a student gets special treatment. I use xtmelogit command to estimate a model with predictors at both level-1 and level-2. 

The model consists of the following variables at level-1: gender (dummy), immigrant (dummy), mothers marital status (dummy), adopted (dummy), class step (categorical), age when beginning in school (interval), parents income (interval), parents education (categorical), parents social status (interval), mothers age (interval), fathers age (interval). The model has one variable at level-2: the ratio of children in the municipality attending private schools. The level-2 identifier is municipality.

After the modeling I have four questions in relation to the xtmelogit:

1) In an xtmelogit model, how can I predict the different expected probabilities of children getting the special treatment in the level-2 municipalities?

2) How do I estimate the probabilities of the predictors in the model, for instance gender or parents income?

3) Can I predict the marginal effects of the variables, and get some kind of a standardized expression of the variables in the model?

4) I want to rank the level-2 objects using the residuals, but what is the interpretation of the residuals at level-2 in an xtmelogit model?  

Thank you in advance for any suggestions/comments.

Soeren Teglgaard Jakobsen
Consultant . MSc (Political Science)
+45 72 26 99 82 . stj@krevi.dk

KREVI . Danish Evaluation Institute for Local Governments
Olof Palmes Allé 19 . DK-8200 Århus N   
post@krevi.dk . +45 72 26 99 70 . www.krevi.dk

////////////////////////////////////////////////////////////

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/